사영 변환하의 데이터 매트릭스 검출

Data Matrix is used in various industrial fields to embed diverse information in a compact fashion. Data Matrix is usually attached by printing method or is marked by laser on the surface of objects. In addition, cameras are used for decoding. It has an Lshape comprised of two perpendicular solid lines that serve to indicate the correct orientation of the code and the boundaries of the data area. Conventional decoding algorithms of Data Matrix require an image acquired under similarity transform. We present an algorithm for the robust detection of Data Matrix under general perspective transform. The presented algorithm first detects the whole area containing the Data Matrix using image binarization, connected component analysis and morphology. Next, corner points corresponding to the L-shape in the Data Matrix are detected using line fitting through polygonal approximation after contour processing. Finally, the Data Matrix is converted into canonical image using homography that is computed using the four detected corner points of the Data Matrix. Experiments using images having large perspective distortions acquired under various pose demonstrates the robustness of presented method.